FIELD METHODS Elicitation Techniques

Read Chapter 12 of Chelliah, Shobhana L., and Willem J. de Reuse (2011) Handbook

of Descriptive Linguistic Fieldwork. Springer: New York

I. Three Questions on Grammar Gathering

Answers to three main questions define a researcher’s position on how and why data is gathered for linguistic analysis. To choose the methods of data gathering, these posi-tions must first be understood. These questions are:

! Should data gathering be theory-driven or data-driven? ! Is data gathering performed in a predictable linear fashion? ! What data should be gathered?

" No linguist is a theoretical “tabula rasa”, nor is it desirable (or even possible) to be one. Thus each fieldwork data-gathering endeavour has some theoretical under-pinning. In addition, theory-driven research can be useful in determining what data to collect, especially with regard to more nuanced grammatical points.

" For the fieldworker, it is better to observe and describe linguistic facts regardless of whether or not they fit into someone’s formal theory. “Paying attention” does not mean that the fieldworker can or should try to pay attention to all morpho-syntactic features of a language at the same time.

" A fieldworker first begins to examine and analyze smaller morphosyntactic struc-tures, and then moves on to examine and analyze larger morphosyntactic struc-tures; however, this does not mean that there is a directionality in fieldwork. There is nothing wrong in adopting this method, as long as one understands that the resulting documentation and description will depend on a simultaneous study of all aspects of the phonology, morphology, and syntax of the language.

II. Data Gathering: How to…

The term “elicitation” is vague. For some researchers, elicitation implies the use of ques-tionnaires and questioning-without-context; it excludes data from texts. For others, el-icitation potentially covers any sort of questioning methodology. Another problem with the term “elicitation” is that it seems to imply data gathering without linguistic analysis. But there is no such thing as data gathering separable from analysis. Some analysis in the field is unavoidable, II.1 Schedule-Controlled Elicitation: The fieldworker has a schedule or questionnaire of material to elicit, prepared by the fieldworker or by another linguist, and asks the ques-tions in the order of the schedule. Elicitation schedules or questionnaires focusing on specific language families or language areas are usually more useful than those covering “the languages of the world". It is important that both the consultant and the field-worker understand that the schedule is just a guideline; it can be modified or even set aside if need be. And even though the fieldworker is the ultimate guide, feedback from

the consultant on how to best proceed with the elicitation is always welcome. " A questionnaire asking the fieldworker analytical questions about the language is

called an “analytical questionnaire”, and sometimes, rather redundantly, “ques-tion questionnaire. Examples of questions on such a questionnaire are “Does the language have a case marking system?” If yes, “Is the case marking system nomina-tive/accusative, or absolutive/ergative?” and so on. These questionnaires, while useful aids to memory, depend entirely on the skill and imaginativeness of the field-worker in eliciting the answers to the questions.

" The most widespread type of questionnaire is the questionnaire with lists of sen-tences to translate into the target language. One problem with most such ques-tionnaires is that they contain sentences out of context

II.2 Analysis-Controlled Elicitation: under this type of elicitation there is no schedule. The analysis (in the most informal sense of the word) controls the elicitation, although it is essential to have some sort of “script” (or “protocol”). (In choosing an approach to an-alysis-controlled elicitation, it is important to take into consideration the degree of ob-solescence or endangerment of the language. As shown by Mithun, there are many sorts of analysis-controlled elicitation that cannot be used effectively when speakers are no longer dominant in the target language.

A. Target Language Interrogation Elicitation: This is elicitation where questions are asked in the target language. One should always record not only an answer specific to the situation, but a variety of appropriate answers as well. This tells the field-worker about the real meaning of the question (Healey 1964:13). A more compli-cated type is to describe a situation and ask what the person might say in that situation at a given moment.

B. Stimulus-Driven Elicitation: In this type, the consultant is provided with some con-crete objects, or pictures, video clips, or a movie, and is asked to comment on them. Several subtypes are distinguished here:

1) PROP-DRIVEN ELICITATION is particularly useful for the study of noun classification The fieldworker will need a number of different types of objects varying in ani-macy, shape, size, surface, consistency, and paired or not paired, as well as objects contained in different ways: in a plate, cup, basket, bottle, or bag.

" One can only talk about objects within sight, ideally objects on the table in front of the fieldworker and consultant.

" The deictic nature of props needs to be taken into account. So the sentence that the fieldworker thinks means ‘I pick up this knife’, might well mean ‘you pick up that knife’.

" Finally, as soon as something slightly abstract is pointed at, misunderstandings will occur. E.g., if the fieldworker tries to get the term for “mirror” by looking into the palm of her hand, she may get as translation something like ‘you are looking at it’, or ‘your hand.’

2) PICTORIAL STIMULUS-DRIVEN ELICITATION involves the use of pictorial stimuli, including line drawings, photographs, and video-clips. In this method, the fieldworker shows a picture to the consultant and asks him/her to describe it, or to comment on it. This method can be used for eliciting quantifier facts (such as the notori-

ously difficult every, some, each, and all, or in the study of spatial or locative rela-tions, verbs of motion and change.

" The consultant might respond from the perspective of the speaker, and not from the perspective of the objects.

" If there are too many pictures on one sheet, or if there are discrepancies in the size of the depictions, consultants will be easily confused.

C. Target Language Manipulation Elicitation: In this type, some word\structure of the target language is manipulated and the consultant has to react to\correct it.

1) WORD-LIST BASED ELICITATION: the fieldworker asks the consultant to make up a sen-tence using the word “X”, selected from a previously collected word list. A draw-back of this method is that the consultant might use a sentence frame that works most of the time.

2) PARADIGMATIC SUBSTITUTION ELICITATION: The fieldworker writes a sentence on a black- board or on paper. Then s/he deletes a word or constituent from the sentence, and asks for another word or constituent that can fit that slot. Asking for another word or constituent is preferable to suggesting one because it avoids the creation of artificial forms by the fieldworker

3) FIELDWORKER-DRIVEN PARADIGMATIC SUBSTITUTION ELICITATION: like Paradigmatic substi-tution elicitation, except that the open slot is filled with another word or constitu-ent by the fieldworker him/ herself. Useful for hard-to-define elements like evi-dentials, modal particles, or other elements that the fieldworker has noticed in conversations and texts. Care must be taken to ensure that just one or two ele-ments of the sentence are changed at a time in order to keep close control over the data.

4) APPROPRIATE ADJUSTMENTS ELICITATION: Useful for fine-grained grammatical analyses like tense, number or person, the fieldworker inserts a novel morpheme or phrase (by either replacing a similar category in the original or adding a novel one) into a sentence, and asks for judgements about grammaticality. If the output is ungram-matical, the fieldworker must note the adjustments that are needed to make it grammatical. One difficulty with this method is that the fieldworker and the con-sultant will have to agree on a transformational metalanguage

5) TRANSFORMATIONAL ELICITATION: In this method, the consultant is asked to transform a given sentence to some specified grammatical target – passive, negative, ques-tion, etc. in the fieldworker-driven version, the fieldworker begins with a sentence in the target language and modifies it in some way, such as moving a constituent. Then the fieldworker asks for a grammaticality judgment, or asks about the prag-matic situation where the “changed” sentence can be used, and asks about any socio-linguistic details. This method can be problematic if the resulting sentences sound too contrived. Also, speakers may have a hard time judging the grammati-cality of the resultant constructions.

6) CORRECTIVE ELICITATION: this method involves the fieldworker deliberately producing an ungrammatical sentence or using an incorrect form to test a hypothesis. The fieldworker produces the form and then waits for the consultant’s reaction. The method is useful in gauging how well the speaker is able to correct the field-

worker. Is the consultant accommodating, uncaring, or afraid to hurt the field-workers’ feelings?

D. Target Language Translation Elicitation: In this method, the fieldworker asks the consultant to translate target language text or other materials gathered during previous sessions.

" Some consultants are very skilful translators, some can offer only a brief or inac-curate synopsis of what the text is about, some can only do word-for-word transla-tions, and others are simply incapable of translation. Therefore the fieldworker needs to know the consultant well enough to be able to judge his/her abilities.

" The consultant might offer ungrammatical English translations to indicate that the equivalent in the language is also ungrammatical. The accuracy and reliability of translations can be affected by differences in the variety of English (or any other contact language) spoken by the fieldworker and the consultant.

" Translation is quite useless in some areas of morphosyntax.

E. Target Language Construction and Introspective Judgment Elicitation: In this type, the fieldworker constructs sentences in the target language. S/he simply relies on the consultants’ intuition and their ability to introspect, and asks them to judge if the constructed sentence is grammatical or not. This is, in simple terms, the “can you say this?” method, and the fieldworker is content with a simple “yes” or “no” answer.

" Grammaticality judgments vary not only across idiolects of different speakers, but even for a single speaker at different points in time. Judgments become blurry. If you are not sure whether a response of ‘yes, it’s fine’ means that the sentence is grammatical or simply that your consultant is being polite, a slightly better ques-tion to ask is ‘would a fluent speaker of the language ever say this?’ This is a good way to weed out sentences that are technically grammatical, but are not preferred. If you get the response ‘well, you could say it’, that almost always means that there is something wrong with the sentence somewhere. Asking when someone might say a sentence like that can also produce interesting responses.

" The notion of ‘grammaticality’ can very easily be confused with the speaker’s no-tion of ‘understandability’. It is not always easy to ensure that this kind of judge-ment corresponds to what people would actually say. So, you should avoid strat-egies such as the following in your elicitation sessions: Is X grammatical? Can you say X? Rather, you should formulate your elicitation sessions with questions more like the following: Would people say X?

" It is also very difficult for you to control for all possible variables when you pres-ent a sentence in isolation for a grammaticality judgement like this. A simple change in intonation—which you may be completely unaware of—may completely change the acceptability of an utterance within any particular real-world context. One way of giving grammaticality judgements greater reliability is to present peo-ple with two (or more) options and to allow them to indicate which they think is the better sentence. Thus, rather than any of the strategies above, it may be a bet-ter idea to ask somebody: Which is better: X or Y?

" Suppose that a completely new use of a morpheme has appeared for the first time, and you want to ask your language-helper: Why did you say X rather than Y? Very of-

ten, the answer to this kind of question calls for much greater linguistic sophisti-cation than any linguistically untrained language-helper is going to be in a posi-tion to provide. Find a way to get at the meaning through contrasting sentences.

" In addition to the fact that introspection is unreliable, there are a variety of other reasons why consultants cannot provide grammaticality judgments reliably. One, consultants aim to please the investigator, they may be guided by prescriptive correctness, factual (in)accuracy

F. Reverse Translation Elicitation: The consultant is asked to translate sentences from the contact language into the target language “How do you say X in your language?”

" Extreme caution is advised here, as most consultants will translate word for word. Even if one is lucky enough to work with a skilled translator, the results can still be odd, unnatural, and/or unidiomatic.

" If one has to use it, do so only late in the fieldwork and preferably through a writ-ten task, so that the consultant has time to think over a written translation

G. Review Elicitation: this procedure is used to check facts already elicited, and there-fore presupposes some method of previous elicitation. The fieldworker reviews previous materials together with the consultant, and looks for agreements, dis-agreements, variants and “mistakes”. As a controlling device, it is useful to ask for a translation back into the target language of something previously elicited

H. Ancillary Elicitation: Also called Text-based elicitation. Texts are crucial for mor-phosyntactic elicitation. In this approach, the fieldworker starts from a previously recorded and translated text (ideally a text of naturally occurring discourse). S/he then elicits more information on the sentences of that text. It is best to start work-ing with shorter sentences. For example, based on a text, one could develop a hy-pothesis about word order and constituent structure. The fieldworker could then change word order to see if it is strict or variable.

" Consultants might have completely different expectations of what the field- worker is expected to understand or not understand in a text. Loanwords from English, which sound assimilated to the fieldworker, might be considered unas-similated by the consultant and therefore obvious to everyone.

" It is not unusual to come across speakers who cannot manipulate a text in any way. Needless to say, these speakers should not be relied upon for ancillary elicita-tion, but they might be helpful with other tasks.

I. Covert Elicitation: In this methodology, it appears that a particular elicitation pro-cedure is being followed, but in fact, the fieldworker is actually paying attention to something else, and the consultant is not aware of this. Covert elicitation can be used when the speaker is bored by a repetitive or technical elicitation session, or when the speaker has strong feelings about what is important and interesting in their language. In such a situation, the fieldworker might try to build the senten-ces for elicitation around linguistic matters that generally interest or amuse the consultant.

" There is an ethical problem here. At worst, the fieldworker can be perceived as de-ceptive, or at best, hurting the consultant’s feelings.

J. Meta-Elicitation: at later stages, once the fieldworker is sure of a particular con-

sultant’s linguistic sophistication, s/he can ask for simple types of analysis. For in-stance, the fieldworker might ask the consultant what the two clauses are that make up the sentence. Or it might be reasonable to ask the consultant directly if there is a semantic difference between two constructions. Usually the consultant cannot tell the fieldworker, but it is worth seeing if s/he has a take on it.

" It is important for the fieldworker to keep analytical questions simple, and not base them too heavily on theoretical concerns.

In a radical departure from the usual tradition in this course, we wil l try and uti l ise as many of the above investigation methods as possible . These are to be conducted in groups of three students each. Before each el ic itation session, groups wil l submit a detai led plan of the method of el ic itation they wil l use for the topic assigned to them for the session.

III. Data Gathering: What to… in MORPHOSYNTAX

As discussed, you will each be individual topics for research. The following list of topics areas must be covered in every individual field report, and will be elicited in class.

Morphology Concatenative morphology: affixation and compounding Nonconcatenative Morphology: Supraseg-mental, Subtractive, Replacive, and Redupli-cative. Inflectional and Derivational morphology

30111.2 Grammatical Typology and Terminology

three-way distinction, e.g., tomorrow/today/yesterday, but there are also many languages that have only one word for both ‘tomorrow’ and ‘yesterday’; that is, a system contrasting ‘today’ with ‘a day away from today’, either plus or minus.

Spatial deixis is further surveyed and studied in Levinson (2003), and Levinson and Wilkins (2006).

It is also useful to distinguish between anaphoric deixis, in which the deictic element (in bold) points backwards, as in … and John said the same thing, from cataphoric deixis, in which the deictic element (in bold) points forwards, as in The winners are the following: .…

Another useful distinction is that between absolute deixis versus relativized deixis. Relativized deixis typically occurs in indirect speech. An example of abso-lute temporal deixis is He was sick, where the past tense of was refers to a time prior to the moment of speech. But in John will say he was sick, the past tense refer-ence of was is not absolute, but relative to the event John will say (Anderson and Keenan 1985:301). The issue of relativized deixis is further discussed in Anderson and Keenan (1985:301–307).

Deixis can be encoded morphologically, through cliticization, or syntactically by the use of Modifiers and Pro-forms.

Grammatical Categories Characteristic of Verb Phrases

The following is a list of the grammatical categories that are characteristically indi-cated on verbs or verb phrases:

TenseAspectMood/Mode/ModalityVoiceValencePolarityControl/NoncontrolSubject person and numberObject person and numberAssociated motion

All of the above can be encoded morphologically, though cliticization, or syntacti-cally by the use of separate words (Adjuncts such as auxiliaries, or Pro-forms) or through constituent order.

Tense

The categories Tense, Aspect, and Mood/Mode/Modality are often intimately connected, and are often discussed together as Tense/Aspect/Mood (TAM) catego-ries. For example, there is considerable overlap of irrealis mood with future tense,

296 11 What to Expect in Morphosyntactic Typology and Terminology

A negator (or a prohibitive, if it is imperative or optative) (assuming it is not an Adverb)A politeness marker (assuming it is not a Verb, Modifier, or Proform)

It is easy to abuse this category, so it is best to limit the number of elements are classified this way. As argued in Section 11.5, one should as much as possible avoid positing a category of “particles”, a wastebasket category for uninflected words.

11.2.3.2 Grammatical Categories

An equally important task that a fieldworker undertakes involves determining the grammatical categories at work in the morphosyntax, morphology, and syntax of a language. Many speakers have an intuitive notion of what a grammatical category is in their language. An English speaker will easily recognize number (singular vs plural); a French or Spanish speaker will easily recognize gender (masculine vs feminine). However, grammatical categories are sometimes not very obvious, and it can be difficult for a fieldworker to determine which ones are relevant in a particular language. In this section, we provide a rough guide to grammatical categories, and some hints on how to identify them.

Grammatical categories are different for noun phrases (Section “Grammatical Categories Characteristic of Noun Phrases”), for verb phrases (Section “Grammatical Categories Characteristic of Verb Phrases”), and for clauses or sentences (Section “Grammatical Categories Characteristic of Clauses or Sentences”).

Grammatical Categories Characteristic of Noun Phrases

The fieldworker should be prepared to look for the following characteristics of nouns or noun phrases:

Semantic RolePragmatically Marked StatusCasePossessionNumberNoun class or genderNoun classificationDeixis

These are discussed in detail below.

Semantic Roles

Semantic roles (or theta-roles, as they are called in the generative literature) were first posited by Gruber (1965) and Fillmore (1968), and are, technically speaking,